AI agents are moving toward a role that goes beyond answering questions or generating text: they are being designed to take actions, choose services and make payments on a user’s behalf. That shift is opening a new debate over the financial networks that could support machine-driven commerce.
Cathie Wood, chief executive of ARK Invest, said investors may increasingly need to “follow the agents” as they assess where technology is heading. She made the comment during a panel at Robinhood’s Summit in Houston on Wednesday, in a discussion that also covered artificial intelligence, private markets and technology investing.
The idea builds on Wood’s earlier advice to “follow the developers” to identify technologies gaining momentum. Developers often adopt tools they find useful, making their choices a possible indicator of future demand. If large numbers of AI agents begin selecting software, data services, computing resources and payment networks independently, their activity could become another signal for investors to watch.
From digital assistants to autonomous spending
AI agents differ from systems that only answer questions or create content. They are intended to complete tasks for users, potentially including arranging services, purchasing digital products or executing other instructions within limits set by the user.
That raises a practical question: how should an agent pay? A user might authorize an agent to spend as much as $500 to book a hotel, for example, without granting unrestricted access to a bank account. The user should also be able to withdraw that permission and review a record of the agent’s actions, according to Joseph Chalom, co-CEO of SharpLink and former head of digital assets at BlackRock.
Chalom has argued that users should not be permanently tied to one bank, technology company or payment provider. In his view, people should be able to move an agent, along with its identity, financial information and permissions, to another provider. He compared that concept with the ability to transfer a phone number between mobile carriers.
The broader concern is control. Chalom wrote that a financial system for intelligent agents would have limited value if a small group of companies determined where users’ money could go. The debate is therefore not only about whether software can spend funds, but also about the authority granted to agents and the infrastructure governing each transaction.
Why crypto networks are part of the debate
Chalom has identified open blockchains such as Ethereum as one possible foundation for agentic finance. A shared network could allow agents, applications and companies to transfer money without every transaction passing through a separate closed system operated by one bank or technology platform.
Stablecoins and blockchain-based payment protocols are also being considered for machine-to-machine transactions. Stablecoins are digital tokens designed to maintain a stable value, while blockchain networks can allow software to send payments directly to other software. Those features could be useful when an agent needs to pay for an application programming interface, purchase data or rent computing capacity without waiting for a person to approve every individual payment.
BlackRock made a similar connection in a September paper on artificial intelligence and digital assets. The asset manager said AI agents could create demand for payment systems designed around machines. Coinbase’s x402 is one example cited in the discussion; it is designed to enable automated payments for online services, including data and API access.
There are also early signs of AI agents appearing elsewhere in crypto markets. Coinbase chief executive Brian Armstrong said on X that Grok was currently the leading client for agentic traders on Coinbase, although he did not provide figures or additional details.
Open networks face established competitors
Crypto is not the only part of the payments industry developing systems for AI-driven purchases. Stripe, Visa, Google and OpenAI are among the companies working on ways for agents to make transactions. BlackRock said traditional payment systems would continue to matter, meaning future agent commerce could be divided between blockchain-based networks and existing financial infrastructure.
That competition gives Wood’s “follow the agents” idea a financial dimension. Investors could monitor not just which AI models and software tools agents use, but also how they pay and which networks handle those payments. Such activity could help show whether stablecoins and blockchains are gaining practical use or whether most agent transactions remain within conventional payment systems.
Moneyiar market picture
Moneyiar market data at the stated times shows mixed performance among the listed assets:
- Monad (MON) was priced at $0.0324794127 at 09:33 on October 5, 2026, down 5.14% over 24 hours but up 16.22% over seven days.
- Aethir (ATH) stood at $0.0074040571 at 09:34, falling 4.74% in 24 hours while gaining 23.92% over seven days.
- RealLink (REAL) was $0.0100747766 at 07:04 on September 15, down 86.56% over 24 hours and 86.35% over seven days.
- ONEchain (ONE) traded at $0.1481029642 at 09:34 on October 5, rising 3.25% over 24 hours but declining 1.48% over seven days.
These figures describe market movements at the supplied timestamps and do not establish a connection between the assets and the development of AI-agent payments.
